Abstract
This paper studies submerged parallel bars as a wave reflector using the
linear Shallow Water Equation. First, we recall the analytical study from
literature about finding an optimal width of a one-bar wave reflector with
a specific height. Numerical study is applied using the Lax-Wendroff dis-
cretization method with different partition along x−axis in order to avoid
numerical diffusion error. Comparison between analytical and numerical
solutions shows a good agreement, qualitatively and quantitatively. Using
an analogous physical argument as in the case of one-bar wave reflector,
we can obtain an optimal dimension of submerged n-bar wave reflector.
Numerical computations are made and they confirm this optimal dimen-
sion. Finally, we conclude that in order to get the minimal amplitude of
transmitted wave, the wavelength of an incident wave and the dimension
of bars should be closely matched.
1 Introduction
When a wave pass over a seabed covered with bars, its amplitude is reduced,
because part of it is reflected. Study of wave propagation past a submerged
parallel bar has many practical applications in coastal and ocean engineering.
This is useful for designing series of bars as breakwaters and bank protection
structures. This subject has been studied by many researchers. Most of them
study sinusoidal beds as wave breakers related to Bragg resonance phenomenon.
See Mei & Liu [2] for a review.
Each time an incoming wave enters a region with a change of depth (which is
significant within its wavelength), it will scatter into a transmitted wave and a
reflected wave. In the case of an incoming wave pass over a bar, i.e. a depth dis-
continuities at x = 0 and x = L, there is a back and forth scattering processes,
each corresponds to a change of depth. Hence, on x > L the transmitted wave
is actually a superposition of several waves that result from several scattering
processes. Generally, the amplitude of transmitted wave At is smaller than the
amplitude of incoming wave A, and it will be minimum if all waves superpose
destructively at x = L. This happens when 2L equals to an odd number times
half wavelength. The dependence of |At /A| with respect to the width of the
bar L is periodic. Meaning that a wider bar is not necessarily a better wave
reflector, and so there is an optimal bar width Lopt . The above result can be
1
found in C.C. Mei [1], see also [3].
with (h0 − h1 ) > 0 denotes the height of the bar. The governing equation
considered here is the linear Shallow Water Equation (linear SWE)
½
ut = −gηx
(2)
ηt = −(h(x)u)x ,
where η(x, t) denotes the surface elevation measured from the undisturbed water
level, and u(x, t) denotes the horizontal component of fluid particle velocity
along the surface. Equations (2) are equivalent to
½
utt = (c2 (x)u)xx
(3)
ηtt = (c2 (x)ηx )x
with c2 (x) = gh(x). For flat bottom hj , the first equation of (3) has the
d’Alembert (monochromatic) solution
η(x, t) = A exp i(kj x − ωt) + cc. + B exp i(kj x + ωt) + cc. (4)
p
with frequency ω that satisfies a dispersion relation ω = kj ghj , j = 0, 1.
2
Imagine that a monochromatic incident wave A exp i(k0 x − ωt) coming from
the left, enters a region with shallower depth h1 . Some energy is transmitted
beyond the step and some is reflected backward, creating scattered waves: a
transmission wave above the depth h1 and a reflection wave above the depth
h0 . When the transmission wave enters an original depth h0 , again it scatters
into two waves. The same also with the reflection wave, it will scatter into two
waves each time it meets bottom discontinuity. Hence, there are infinitely many
scattering processes. Let we assume the general solution η(x, t) of (3) is
η(x, t) = exp(−iωt)ν(x) (5)
where
A exp ik0 x + Ar exp −ik0 x, x<0
ν(x) = a exp ik1 x + b exp −ik1 x, 0≤x<L (6)
At exp ik0 (x − L), L ≤ x,
with Ar , a, b, At have to be determined. Demanding the surface profile η to be
continuous at the jump x = 0 and x = L implies that ν(x) should be continuous
at x = 0 and x = L. Another jump condition will be obtain in the following.
Substituting η in (5) into the second equation in (3) yields
−ω 2 ν(x) = (c2 νx )x .
Integrating the equation above with respect to x from x = 0− to x = 0+ , and
using the condition that ν(x) is continuous at x = 0, we obtain
lim h0 νx = lim h1 νx . (7)
x→0− x→0+
Hence, the surface profile η(x, t) = exp(−iωt)ν(x) for a bottom with a step is
given by (6) with Ar , a, b, At must be found by matching ν(x) and h(x)νx (x) at
the two edges. Some calculations (see [1] for details) will result
¯ ¯
¯ At ¯ S 2 − D2
¯ ¯= (8)
¯ A ¯ |S 2 E −1 − D2 E|
¯ ¯
¯ Ar ¯ −1
¯ ¯ = |SD(E − E)| , (9)
¯ A ¯ |S 2 E −1 − D2 E|
with
S = h1 k1 + h0 k0 , D = h1 k1 − h0 k0 , E = exp ik1 L. (10)
2 2
Further, we can also show that |At | + |Ar | = |A|2 , which is energy conserva-
tion. Note that the above results also holds analogously for a wave entering a
trough, i.e. h0 < h1 .
The curves of |At /A| and |Ar /A| as periodic functions of L are shown on Figure
1. Let Lopt denotes the minimal width of a bar that gives minimal |At /A|, hence
π p
Lopt = 1/4λ1 = gh1 . (13)
2ω
3
√ √ √ √
Since we can rewrite S = √ωg ( h1 + h0 ) and D = √ωg ( h1 − h0 ), then
formula (11) can be written explicitly as a function of h1 /h0 , as follows
¯ ¯ √ √ 2 √ √ 2 p
¯ At ¯ ( h 1 + h 0 ) − ( h 1 − h 0 ) 2 h1 /h0
¯ ¯ = √ √ 2 √ √ 2 = (14)
¯A¯ ( h1 + h0 ) + ( h1 − h0 ) (1 + h1 /h0 )
min
The plot of the curve |At /A| as a monotonically increasing function of h1 /h0
for L = Lopt is given in Figure 2. Note that |At /A| = 1 when h1 = h0 , and
|At /A| = 0 when h1 = 0, as we expect.
Figure 1: The periodic curves of |At /A| (above) and |Ar /A| (below) as functions
of L together with their values obtained numerically (dotted). Here, we use
h0 = 10 and h1 = 0.4 × h0 .
Figure 2: The curve |At /A| as a function of h1 /h0 together with its values
obtained numerically (dotted), when the bar width is L = Lopt = 41 λ1 .
4
3 Discretization and Simulation
Consider the linear Shallow Water Equation (2) with a piecewise flat bottom
(1) in a domain (x, t) ∈ [a, b] × [0, T ], with a < 0 < L < b. The discretization
that will be used is the Lax-Wendroff method, which is of order O(4x2 , 4t2 )
in accuracy. The von Neumann stability analysis as in [4] yields the Courant
³ ´2
4t
number Cj ≡ ghj 4x should be less than one. Numerical diffusion error will
not appear if and only if Cj = 1, for j = 0, 1. In this numerical discretization,
numerical diffusion should be avoided completely in order √ to make a quantitative
comparison.√Hence, for a certain 4t, we take 4x1 = gh1 4t for 0 < x < L,
and 4x0 = gh0 4t, elsewhere. Hence, the descritized linear SWE equation for
each flat depth hj , j = 0, 1 reads
³ ´2
ηin+1 = ηin − hj 24x 4t
(un
− u n
) + 1
gh 4t n
(ηi+1 − 2ηin + ηi−1
n
)
j i+1 i−1 2 j 4xj
³ ´2
un+1 = un − g 4t (η n − η n ) + 1 ghj 4t
i i 24xj i+1 i−1 2 4xj (uni+1 − 2uni + uni−1 ).
(15)
We will simulate a monochromatic wave coming from the left, enters a domain
with a submerged bar. So, we use the still-water initial condition η(x, 0) = 0
and
q u(x, 0) = 0 and the left boundary condition η(a, t) = A sin ωt and u(a, t) =
g
h0 η(a, t). Note that these boundary conditions should be satisfied by the
right running wave solution of the linear SWE (2). For the right boundary
u(b, t) and η(b, t), we apply the Forward Time Backward Space method of the
order O(4x, 4t), that yields a right absorbing boundary condition. This lower
order accuracy will directly leave the computation domain following the right
going characteristic, and will not effect the higher accuracy O(4x2 , 4t2 ) in the
interior domain. As a check, simulation for flat bottom h0 using the above ini-
tial and boundary condition will show a monochromatic
√ wave with amplitude
A running to the right with velocity c0 = gh0 , as we expect.
For simulation we use spatial interval [0, 150], time interval [0, 22], gravitation
g = 10, frequency ω = 1, and a bottom topography
½
4, for 50 < x < 50 + (Lopt ≈ 10)
h(x) = (16)
10, for others,
Figure 3 shows the result of numerical computation when a monochromatic
wave running above topography (16). It clearly shows that a wave scatters
into a transmission wave running to the right and a reflected waves running
to the left. The numerical values of |At /A| and |Ar /A| are given in Figure 1.
We hardly can see the difference between the analytical and numerical results.
For the choices of 4t = 0.01 the results agree up to O(10−4 ) as expected. So
we can conclude that numerical computation also admits energy conservation
|At |2 + |Ar |2 = |A|2 .
5
Figure 3: The surface wave profiles η(x, t) at several times. The grey area
indicates the position of a submerged bar.
ize the physical argument in Section 2 about destructive interference that result
the smallest net amplitude of the total right-going wave in x > L. Imagine
an incidence wave above a flat depth h0 , running through the first bar with
depth h1 . When the width of the first bar L1 = (n + 12 ) 12 λ1 , n = 0, 1, · · ·, the
amplitude of the total right going wave in x > L1 is minimum, and from (11)
it equals to (S 2 − D2 )/(S 2 + D2 )A. The back and fort process of transmitted
and reflected waves in L1 < x < L1 + L0 will also give a minimal net amplitude
of total right going wave in x > L1 + L0 when L0 = (n + 21 ) 21 λ0 , n = 0, 1, · · ·.
The minimal amplitude is now (S 2 − D2 )2 /(S 2 + D2 )2 A. The same thing, the
width of the second bar must be L1 = (n + 12 ) 12 λ1 , n = 0, 1, · · ·, in order to get
a minimal amplitude of transmission wave in x > (L0 + 2L1 ). And the minimal
transmitted wave amplitude is now (S 2 − D2 )3 /(S 2 + D2 )3 A. Hence, the most
optimal 2-bar wave reflector with height h0 − h1 has width L1 opt = 41 λ1 , and
the two bars are separated at distance L0opt = 14 λ0 .
Several numerical computations finding |At /A| using L0 opt , for several values of
L1 is made, in order to show that the optimal width is L1 opt = 14 λ1 . The result
is shown on Figure 4. The periodic curve of
¯ ¯ µ ¶3
¯ At ¯ S 2 − D2
¯ ¯= (18)
¯A¯ |S 2 E −1 − D2 E|
as function of L1 for L0 opt is also plotted. Hence, the identical 2-bar wave re-
flector height h0 − h1 = 0.6 × h0 with optimal width will reduce the amplitude
up to |At |min = (S 2 −D2 )3 /(S 2 +D2 )3 A or 69% of the incoming wave amplitude.
In the case of n−bar wave reflector, following the physical argument we can
easily find its optimal dimension. The main point is that above each bar, and
above a space between the two adjacent bars, all waves should interfere destruc-
tively. The destructive interference will be obtained when the width of each
6
Figure 4: The periodic curve of |At /A| as function of L1 together with their
values obtained numerically
√ (dotted). We use h0 = 10, h1 = 0.4 × h0 and
L0√opt = 14 λ0 = π2 gh0 ≈ 15.7. This confirms the optimal value L1 opt = 41 λ1 =
π
2 gh1 ≈ 10.
piece of the piecewise constant depth function is one forth of the wavelength,
or formula (13). And the transmitted wave amplitude can be obtained analo-
gous to formula (18). In the case h1 = 0.4 × h0 , numerical calculations for the
three-bar and four bar wave reflector reduce the amplitude up to 74% and 60%,
respectively. The above argument can be generalized directly to get the optimal
dimension of the nonidentical n−bar wave reflector.
Further, we will relate this optimal dimension with Bragg resonance. This phe-
nomenon happens when a monochromatic wave enters a region of sinusoidal
bar with wavelength half of the wavelength of the incident wave. When this
phenomenon occurs, the amplitude of the reflected wave is maximal, and the
amplitude of the transmitted wave is minimal. Suppose that on a flat bot-
tom with depth h0 there is a series of pbar and trough with amplitude εDh0 .
π
The optimal length of this bar is 2ω gh0 (1 − εD) and length of the trough
π
p
is 2ω gh0 (1 + εD). So, the wavelength of this periodic bar and trough is
π
√ ¡√ √ ¢
2ω√ gh0 1 − εD + 1 + εD . In the limit εD → 0, the wavelength tends to
π
ω gh0 which is a half times the wavelength of the incident wave. This is exactly
the Bragg resonance condition.
5 Conclusions
The Lax-Wendroff discretization method with different partition along x−axis
was applied to solve the linear Shallow Water Equation with a bottom topog-
raphy of a submerged n−bar. Numerical computations confirmed the optimal
dimension of the one-bar wave reflector: the width of the bar should be one forth
times the wavelength of the incident wave (obtained analytically by C.C.Mei in
[1]). The optimal dimension of a wave reflector that consists of a submerged
nonidentical n-bar and the formula of the transmitted wave amplitude were also
obtained. Numerical computations were used to confirm this.
7
References
[1] Mei, C.C., 1982 The Applied Dynamics of Ocean Surface Waves, John Wiley
& Sons.
[2] Mei, C.C., & Liu, P. L-F., 1993 Surface Waves and Coastal Dynamics,
Ann.Rev. Fluid Mech. 25, 215-240.
[3] Groesen, E. van, Andonowati, 2002 Similarities Between Optical and Surface
Water Waves, MIHMI, vol. 8, no. 3. 1-8.
[4] Hoffman, J.D., 1992, Numerical Methods for Engineers and Scientists,
McGraw-Hill, Inc.
[5] S.R. Pudjaprasetya, A.Y. Gunawan, V. Noviantri, G. Siregar, 2008, Can
Longshore Sandbars Act as Wave Reflectors? in preparation.
[6] J., Yu, C.C., Mei, Do longshore bars shelter the shore?, J. Fluid Mech. 404,
pp. 251-268, 2000.